Exercise and Manual Therapy for Diabetic Peripheral Neuropathy: A Systematic Review

: (1) Background: Diabetic peripheral neuropathy (DPN) is one of the most common complications of diabetes mellitus (DM). Control of hyperglycaemia as well as surgical decompression are effective treatments for these patients. However, surgery is not indicated for all candidates. Manual therapy and physical exercise have been shown to be effective for peripheral neuropathies, and exercise for DM. The aim is to review the effectiveness of manual therapy and/or exercise in patients with DPN. (2) Methods: Randomised controlled clinical trials comparing the effects of manual therapy and/or exercise on pain, function and/or balance were selected. The search strategy was performed in PubMed, PEDro, Scopus, Cochrane and Web of Science databases. The PRISMA statement was followed. (3) Results: A total of 656 articles were registered, and 29 were selected. DPN criteria selection. Aerobic, strength and balance exercises are beneﬁcial for DPN. Sessions of 30–60 min, three times per week for 8 weeks seems to be the most used dose. Manual therapy is effective in the short term. A combination of both modalities was more beneﬁcial than alone in one study. (4) Conclusions: Exercise and manual therapy are beneﬁcial for patients with DPN. More studies should be carried out for analysing the potential effect of combining manual therapy and exercise.


Introduction
Diabetes mellitus is a heterogeneous metabolic disease characterised by hyperglycaemia as a result of defects in insulin secretion, in insulin action or in both [1]. It is one of the most common chronic diseases, with an estimated prevalence of 366 million patients in 2030 [2]. Peripheral neuropathy is one of the most common complications of diabetic patients, with 25-50% being affected, especially with type 2 diabetes [3][4][5]. Peripheral neuropathy is known as the injury of small-or large-diameter nerve fibres of the peripheral nervous system, resulting in altered motor, sensory, vibration and proprioception functions for large fibres, and pain, temperature and autonomic function for small fibres [6].
The pathophysiology of diabetic neuropathy is multifactorial, with influence from genetic, environmental, behavioural, metabolic, neurotrophic and even vascular factors [7]. The potential mechanisms of the nerve lesions in diabetes include hyperglycaemia (toxic/reactive metabolites stemming from the hyperglycaemia), microangiopathy and ischemia, cell signalling anomalies due to diacylglycerol and to protein kinase C, sodium channel deregulation and demyelination [8]. Hyperglycaemia is the main factor of risk in the various types of diabetes, provoking lesions by microvascular and metabolic alterations. These vascular changes damage the primary sensory nerves through neuronal hypoxia and nutrient deficits [8,9]. The neural lesion occurs in both long and short fibres indiscriminately, with a differentiated clinical presentation in both cases. The most frequent alterations are produced in short fibres.
There is evidence that strict blood glucose control is effective in halting progression to diabetic neuropathy [10], as it controls the hyperglycaemic component of these patients. Invasive treatment, such as surgical decompression, can be useful for relieving neuropathic symptoms in some patients. However, not all diabetic patients with neuropathy are candidates for nerve decompression [4]. Conservative treatment such as exercising, receiving transcutaneous electrical nerve stimulation (TENS) and taking vitamin D supplements have been shown to be beneficial in reducing pain and improving function and balance in these patients [3,11,12].
During exercise, muscle contraction increases blood glucose capture to complement intramuscular glycogenolysis. To replenish glycogen stores, resting muscle captures glucose postprandially depending on circulating insulin. After exercise, both pathways increase glucose capture towards the muscle [13]. It has also been observed that exercise improves blood glucose control and quality of life, reduces cardiovascular risk and contributes to weight loss [14]. Consequently, exercise-based interventions are beneficial for patients with diabetic neuropathy.
Various systematic revisions have indicated that manual therapy is beneficial in improving function and symptoms in patients with peripheral neuropathies [15][16][17]. In a study with rats, Zhu et al. [4] observed the impact of these oscillations on diabetic neuropathy using mobilisations of the sciatic nerve. The authors found an attenuation of allodynia in the treatment area and a reduction in pro-inflammatory mediator concentration in the sciatic nerve branches.
In spite of the elevated prevalence of peripheral neuropathy and its major involvement in the morbidity of patients with diabetes, it is the chronic complication for which the least information is available about its pathogeny and epidemiology, and the one with the fewest standardised study methods. Additionally, no systematic review on its therapeutic management, including manual therapy and exercise, was found [9]. The aim of this study was to review the effectiveness of manual therapy and/or exercise in patients with diabetic neuropathy. Web of Science "Diabetes Mellitus" AND ("Diabetes Complications" OR "Peripheral Nervous System Diseases" OR "Diabetic Neuropathies") AND ("Musculoskeletal Manipulations" OR "Therapy, Soft Tissue" OR "Manual therapy" OR "Physical Therapy" OR "Resistance Training" OR "Exercise Therapy") -Document type: Article. -Language: English, Spanish, French, Italian.

Eligibility Criteria
Inclusion criteria were: (1) controlled clinical trials that compared a manual therapy and/or exercise group with a control group and/or placebo group and/or other conservative treatment; (2) studies in which the sample included patients diagnosed with diabetic neuropathy; (3) studies that focused on pain, sensory and motor function, and balance outcomes; and (4) studies published in English, French, Italian or Spanish. Exclusion criteria were studies for which (1) surgery was contemplated as a selection and/or intervention process; or (2) another approach was used, except if it was in the context of comparing it with manual therapy or exercise techniques. complete text, following the eligibility criteria. If there were any discrepancies in opinions or any doubts, a third author (C.H.-G.) was consulted to resolve them.

Risk of Bias in Individual Studies
The Physiotherapy Evidence Database (PEDro) scale and Cochrane's Risk of Bias 2 (RoB 2) tool were used to evaluate the methodological quality of the clinical trials. The first instrument, the PEDro Scale, conforms to the Delphi list, based on the consensus of experts to help the reader to identify clinical trials with sufficient internal validity and statistical information to make their results interpretable. It consists of 11 criteria to answer, using "Yes" or "No". Each criteria fulfilled receives 1 point, if the information is clearly expressed in the study; the maximum score is 11 points [19].
The RoB 2 tool is Cochrane's second version for evaluating the risk of bias in clinical trials. The biases are assessed in 5 different domains, as follows: (1) randomisation process; (2) effect of assignment to the interventions; (3) data from the outcome; (4) outcome measurement; and (5) results reported. Within each domain, there are 1 or more questions to which to respond. The answers lead to judgements of "low risk of bias", "some concerns" or "high risk of bias" [20]. The tables and charts presenting the results obtained with RoB 2 were created using the risk-of-bias visualisation (robvis) tool [21].
The first analysis focused on article titles and abstracts, with 492 being excluded for study design, sample, lack of complete article and interventions other than manual therapy and/or exercise.
The 72 articles remaining were then analysed to select those that fulfilled the inclusion criteria. Of these, 29 articles were selected for analysis. The selection process is shown in a PRISMA diagram flow ( Figure 1).

Sample
The study characteristics and sample selection are described in Tables 2 and 3. All the articles offer a sample of 1476 patients. There were 14 articles that specified that the diabetes had to be type 2 [22][23][24][25][26][27][28][29][30][31][32][33][34][35]; the rest of the articles did not specify which type of diabetes should be selected for the sample.
The 72 articles remaining were then analysed to select those that fulfilled the inclusion criteria. Of these, 29 articles were selected for analysis. The selection process is shown in a PRISMA diagram flow ( Figure 1).

Sample
The study characteristics and sample selection are described in Tables 2 and 3. All the articles offer a sample of 1476 patients. There were 14 articles that specified that the diabetes had to be type 2 [22][23][24][25][26][27][28][29][30][31][32][33][34][35]; the rest of the articles did not specify which type of diabetes should be selected for the sample.
Blood analyses taking HbA1c as a reference (cut-off value ≥ 6.5%) and blood glucose levels (cut-off value ≥ 7.0 mmol/L or ≥120 mg/dL fasting value and ≥11.1 mmol/L or ≥200 mg/dL postprandial value); the Michigan questionnaire for neuropathies; and the American Diabetes Association criteria [50] were the criteria most used for the diagnosis of diabetic neuropathy. Three studies included an electrodiagnostic test to confirm the presence of neuropathy [25,35,45]. Other criteria observed in the different studies were physical examination tests, vibratory threshold, sensitivity to fine touch and physical skills.
Moreover, the typology of diabetic neuropathy is specific in a few articles, being described as peripheral neuropathy.
Moreover, the typology of diabetic neuropathy is specific in a few articles, being described as peripheral neuropathy.

Exercise
The majority of the studies compared an exercise group with a control group, implementing patient education or standard care. Four studies included three groups in their analysis: two intervention groups and one control group [23,34,37,47]. The intervention groups were mainly different proposals for types of exercise. The exception was Serry et al. [34], who compared exercise with TENS. Ten studies considered another type of training as a control group or considered two intervention groups directly [23,[30][31][32]34,36,37,47].                   The principal forms of exercise were balance work, aerobic exercise and strength exercises. There were 12 articles that developed various programmes to work on balance [26,27,[29][30][31][36][37][38]40,41,47,50]. In this group, the studies included programmes focused on balance, sensory-motor exercises, gait exercises, bio-feedback and unstable platforms. Seven articles developed aerobic-type training [23,28,34,39,42,46,49]. These exercises were generally carried out at medium intensity using a treadmill or stationary bike. Lastly, four articles used strength exercises as the basis of their training [25,26,32,33]. The strength exercises were differentiated between using body weight or not, or strengthresistance training.
The programmes focused on balance were compared with control groups with patient education and/or routine care, and statistically significant intra-and intergroup improvements were shown. In addition, the training that included vibration showed statistically significant differences as compared with a strength programme [38] or programmes without vibration [37]. Jannu et al. [30] added the wobble board against a balance programme, but they did not find statistically significant improvements in that group. This study used physiotherapy in both interventions. Training with a mini trampoline also observed statistically significant differences compared with a control group [50].
Aerobic training was also compared mainly with control groups; statistically significant differences in favour of the intervention group were found [28,42,46,49]. Cox et al. [23] observed that an exercise programme of high-intensity interval training (HIIT) and moderateintensity continuous training (MICT) showed improvement and significant differences in pain as compared with a control group. However, HIIT training might involve adverse effects if it exceeds 100 h of training. Turning to another focus, it seems that treadmill and stationary bike training is more effective than a balance training group [39]. Lastly, Serry et al. [34] observed that both aerobic exercise and electrotherapy improved pain significantly.
Essentially, strength training was compared with a control group. Mueller et al. [32], however, compared body-weight strength training with another strength training group without body-weight strength. They found statistically significant differences in the variable HbA1c in favour of the group with body weight, in the distance covered in the 6MWT test, and in favour of the group without body weight. The rest of the articles used various methodologies for strength training: isometrics [33], isokinetic exercises [25] and combined with balance [26]. All of these were effective in the balance and strength variables compared with the control group. Furthermore, the study that used isokinetic exercises (in addition) included a programme focused on stability and aerobic capacity.

Manual Therapy
Only six articles included manual therapy as the intervention for patients with diabetic neuropathy. Once again, these studies compared manual therapy with control groups, which were defined as patient education, standard care or no intervention.
Thai-type massage [22], Gua Sha [43], reflexology [48] and aromatherapy [45] seem to be more effective in pain and function variables than the control group in patients with diabetic neuropathy. Treatment using neurodynamics yielded statistically significant favourable changes in the vibratory threshold in comparison with no intervention in these patients [24]. Lastly, the combination of articular, neural and soft tissue mobilisation techniques seemed to be more effective than an electrotherapy protocol in patients with diabetes having carpal tunnel syndrome [45].
In the case of manual therapy, the variables most used in the studies were pain (by a VAS scale) in three articles [44,45,48]; vibratory threshold in three articles [24,43,48]; and other variables linked to function such as questionnaires, sensitivity assessment or range of motion.

Exercise and Manual Therapy
Just one study [35] analysed the effect of an intervention with aquatic exercise, with massage and the combination of these two techniques against a control group (no intervention). Shourabi et al. observed statistically significant improvements in all the groups. In patients with diabetic neuropathy, the combination of aquatic exercise and massage was more effective for the variables related to balance and physiological factors.

Dosage
In the exercise group, the time used for the exercise programmes ranged from 40 to 60 min. Most of the studies integrated warm-up and cool-down phases (from 5 to 10 min for each phase) in the intervention process. The total training period was some 60 min of intervention, principally in the aerobic and balance exercises. In contrast, the programmes focused on strength involved from 20 to 25 min of training.
As for hours a week, the majority of the protocols were given three times a week. The greatest number of sessions was once a day, in the study of Richardson et al. [36]. Dixit et al. [28] applied their intervention from 3 to 6 times a week. The minimum number of sessions per week (once a week) was stipulated in two articles [26,47].
In the manual therapy group, the intervention period stipulated for all the studies was 30 min. Only Singh et al. [24], with 10 min for the neurodynamic techniques, and Xie et al. [43], with 60 min, differed from this period.
Treatment application varied among the articles. Treatment was given three times a week in three articles [22,44,45]; once a day in two studies [24,43]; and once a week in one article [48].
All of the studies carried out short-term follow-ups, from approximately 2 to 4 weeks. However, Dalal et al. [48] had a more long-term follow-up of 5 months.
The study of Shourabi et al. [35], which incorporated manual therapy and exercise groups into their study, also used the predominant doses in the rest of the studies in each speciality.

Evaluation of Methodological Quality
The RoB 2 tools show that the features with the worst methodological quality in the set of studies are the biases in measurement of the variable results, due to deviation from interventions, with approximately 25% being high risk. The biases from the lack of data on the variable results and from the selection of the results reported are the domains with the best methodological quality in the set of studies, around 75% having a low risk of bias ( Figure 2). groups into their study, also used the predominant doses in the rest of the studies in each speciality.

Evaluation of Methodological Quality
The RoB 2 tools show that the features with the worst methodological quality in the set of studies are the biases in measurement of the variable results, due to deviation from interventions, with approximately 25% being high risk. The biases from the lack of data on the variable results and from the selection of the results reported are the domains with the best methodological quality in the set of studies, around 75% having a low risk of bias ( Figure 2). The quality measured in all the studies analysed using the PEDro scale was 6.6 out of 11. Ten of the 29 studies analysed obtained an overall score of low quality [24,27,30,[33][34][35][36]39,40,50], 10 articles received a moderate overall score [23,28,37,38,44,[47][48][49] and 9 articles received a high overall score using this scale [22,25,29,31,32,[41][42][43]45,46] (Table 4). Some of the studies did not blind the subjects and the therapists that administered the therapy. The quality measured in all the studies analysed using the PEDro scale was 6.6 out of 11. Ten of the 29 studies analysed obtained an overall score of low quality [24,27,30,33-36, 39,40,50], 10 articles received a moderate overall score [23,28,37,38,44,[47][48][49] and 9 articles received a high overall score using this scale [22,25,29,31,32,[41][42][43]45,46] (Table 4). Some of the studies did not blind the subjects and the therapists that administered the therapy.

Discussion
The objective of this systematic review was to review the effectiveness of manual therapy and exercise in patients with diabetic neuropathy. We found 29 articles, of which just one analysed the combination of both techniques. There were six articles that analysed the effect of manual therapy, while the rest analysed a variety of exercise programmes. We found a lack of studies that analysed the combination of manual therapy and exercise and only manual therapy in patients with diabetic neuropathy.
The selection criteria were heterogeneous among the studies. Beginning with sample age, the articles differed in the minimum age for sample inclusion, from 18 to 40 or 45 years. The mean age ranges were between 45 and 85 years old. In other reviews, the mean ages of the patients were between 50 and 60 years old [10,12]. This can be explained by the elevated diagnosis of diabetic neuropathy in the 51-60-year age group, in patients that have type 2 diabetes [51,52]. In addition, type 2 diabetes represents the greatest proportion of the total prevalence of diabetes (90%) and is characteristic of adult patients [53].
In line with the previous point, a great number of the analysed studies specified that the patients had to have type 2 diabetes, excluding patients with type 1 from the sample. The studies that include these patients do not provide results based on the type of diabetes that the patients have. Consequently, it is impossible to determine the effectiveness of these treatments in those patients and on diabetic neuropathy.
Moreover, the typology of diabetic neuropathy was defined as diabetic peripheral neuropathy, not considering other types of diabetic neuropathy such as diabetic autonomic neuropathy, cardiac autonomic neuropathy, gastrointestinal neuropathy and genitourinary disturbances. These profiles should also be well identified in the selection criteria because of the problems that exercise can have on patients [14].
Another point to be mentioned is the heterogeneity of the criteria used to diagnose diabetic neuropathy. The articles describe the use of the Michigan Diabetic Neuropathy (MDN) questionnaire, medical diagnoses, sensitivity tests and blood parameters (HbA1c and blood glycaemia). Nagpal et al. [54] observed that the predictive models for diagnosing diabetic neuropathy were ambiguous; electrodiagnostic studies do not enable detecting lesions in fibres of small diameter (which are the most damaged in this type of patient); imaging tests present different limitations and their usefulness has not yet been demonstrated in this type of clinical picture; and biomarkers need further studies [55]. Consequently, no agreement has been reached as to which criterion or criteria should be taken into account in diagnosing diabetic neuropathy. The studies should unify criteria so as to obtain a more representative sample that consider the alterations at the level of both large-and small-diameter nerve fibres.
We have found considerable variability in the treatment with exercise. This variability exists in both the method of exercise and in the analysed variables. All the forms of treatment (aerobic exercise, balance therapy and strength-resistance treatment, as well as the combination among them) seem to be effective in improving the signs and symptoms of patients with diabetic neuropathy. Aerobic exercise, described principally as walking or using a stationary bike at moderate intensity, significantly improved all the variables analysed in the studies. Statistically significant differences were shown against control groups of standard care and/or patient education, but these results were not observed with groups using TENS interventions. Kiani et al. [39] showed improvements using several questionnaires in relation to a group treated with balance-focused exercise. In their systematic review, Gu et al. [56] also found that moderate-intensity aerobic exercise positively impacted neural function in patients having type 2 diabetes with diabetic neuropathy.
Balance programmes were also developed in a very heterogeneous way. The programmes using vibration, biofeedback, Tai-Chi and sensory-motor techniques seemed to have significant improvements after the exercise programme and to have the best benefit compared with control groups of standard care and/or education. Jannu et al. [30] observed that using the wobble board did not yield any benefits against a stability programme. A biofeedback protocol [31] demonstrated significant improvements compared with a multimodal treatment using tendon gliding, resistance training and patient care. The programmes that used vibration also had statistically significant improvements in functional variables as compared with standard strength or balance exercises. In their systematic review, Ites et al. [57] found that the intervention using exercise centred on the lower limb was recommendable, for clinical use, in treating balance dysfunction in diabetic neuropathy. However, they stressed that there were few high-quality studies on this research area.
In our review, we found that the studies using this type of programme were of moderate to high quality. Using these programmes makes it possible to conserve motor control and balance (in spite of sensory deficit) and prevent falls or hyperpressure areas, avoiding ulcerous processes and amputations of the feet. However, just a single study [30] included conventional physiotherapy as a complement to balance training, evading the effects that flexibility or postural re-education programmes can implement to structurally balance the supports, not only at the sensitive level.
The analysed strengthening programmes also yielded benefits in the signs and symptoms of patients with diabetic neuropathy. Although the evidence in this sense is more scarce, this type of exercise is beneficial (whether in isolation or in combination with the two previous types of exercise) in balance and function variables. The American Diabetes Association [14] encourages including strength training for these patients to mitigate neural symptoms and conserve muscle mass in the elderly, as well as to improve quality of life and hyperglycaemia control.
Exercise, in all its modalities, is beneficial for patients with diabetic neuropathy, with high methodological quality of the studies included. However, the variety of the modes used makes it impossible to identify the most beneficial type of exercise for patients with diabetic neuropathy. The American Diabetes Association [14] recommends doing a minimum of 150 min/week of moderate-intensity aerobic exercise or getting 75 min/week of vigorous aerobics, at a minimum of three times a week with a rest between sessions of no more than two consecutive days; 2-3 sessions/week of non-consecutive resistance exercises; and flexibility and balance training 2-3 times/week. Of the studies included in this systematic review, only five [22,27,33,39,40] reached these recommendations. Furthermore, the variety of doses used and of follow-ups make it impossible to specify which type of dosage is the most beneficial for the patients. It would be interesting to design studies based on the recommendations established by the ADA, as well as monitoring cardiovascular parameters (altered in this collective) [14].
As for manual therapy, there seems to be a short-term effect of the different types of massage after treatment in comparison with standard-care and treatment-free control groups, with high methodological quality of the studies. In the case of neurodynamics, improvements are also observed after treatment at the vibratory threshold. Lastly, Talebi et al. [44] observed that an intervention involving articular, neural and soft tissue presented greater benefits than using TENS in patients with diabetes and carpal tunnel syndrome, with moderate methodological quality. It should be pointed out that the only study with medium-term follow-up (5 months) that observed benefits as compared with conventional drug treatment was the study that was of the highest methodological quality. As Zhu et al. [58] observed in a study on rats with diabetic neuropathy, using neural techniques acted on the symptoms and reduced pro-inflammatory cytokines. In addition, it was observed that patients with type 1 diabetes present increased synthesis of type III collagen, which reflects the deposit of matrix and connective tissue in the basal membrane [59]. Manual therapy by local intervention may help to control the symptoms. It has been shown to act at the peripheral, spinal and supra-spinal levels [60]. The mechanical stimuli that such treatment provides favours the microcirculation in tissues, even in nerves. Approaching all the surrounding tissues, such as the joint, soft tissue and even the nerve itself, is more effective against the symptoms because of the increased oxygenation of all the tissues [61].
Shourabi et al. [35] were the only ones to study the combination of the two modalities. It was seen that the combination of both was more beneficial than aquatic exercise and massage alone. However, these results cannot be compared with those of the rest of the studies due to the different variables evaluated and the moderate quality of the study. Lastly, due to the limited number of studies that compared both modalities and the diversity of result variables used to quantify the effects of the techniques, it is impossible for us to conclude whether the combination of manual therapy and exercise is the most effective for the treatment of patients with diabetic neuropathy.
The results of this systematic review have some caveats due to the limitations of the studies included. The lack of homogeneity in the sample selection criteria, along with the variety of result variables used, make it difficult to compare the studies. Special mention should be made of the lack of specific reference in the studies to the procedures carried out for the detection in patients of the possible presence of cardiovascular autonomic neuropathy. Only two studies [39,41] established it as an exclusion criterion, and did not specify the diagnostic criteria used for it. Neither did they include the presence of ischemic heart disease in the exclusion criteria. Examination for cardiovascular autonomic neuropathy is essential before prescribing exercise because it can be asymptomatic and detected only by the lack of variability of the heart rate with respiration, and it is associated with an increased cardiovascular risk [14].
The diversity of exercise modalities and manual techniques also makes objective comparison of the results difficult. The review itself is subject to limitations stemming from the wide range of result variables included for analysis.

Conclusions
In short, the various modes of exercise and manual therapy are beneficial for patients with diabetic peripheral neuropathy. Regarding the combination of both therapies, no conclusion on this hypothesis can be drawn due to the lack of evidence available. Exercise has been widely studied in its different modalities, but new revisions by modes of exercise, with similar result variables, should be carried out to establish more specific protocols. As for manual therapy, the limited number of studies on these patients must be increased in order to obtain greater knowledge about its effect. Finally, in the face of the benefits from both techniques separately, increasing the amount of evidence on the effect of combining them is needed for verifying its potential effectiveness for these patients.